xiaogpt 2.82__tar.gz → 2.84__tar.gz

{xiaogpt-2.82 → xiaogpt-2.84}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: xiaogpt
-Version: 2.82
+Version: 2.84
 Summary: Play ChatGPT or other LLM with xiaomi AI speaker
 Author-Email: yihong0618 <zouzou0208@gmail.com>
 License: MIT
@@ -25,6 +25,7 @@ Requires-Dist: tetos>=0.2.1
 Requires-Dist: groq>=0.5.0
 Requires-Dist: pyyaml>=6.0.1
 Requires-Dist: langchain-community>=0.0.38
+Requires-Dist: lingua-language-detector>=2.0.2; python_version < "3.13"
 Requires-Dist: aiohttp==3.9.5; extra == "locked"
 Requires-Dist: aiosignal==1.3.1; extra == "locked"
 Requires-Dist: annotated-types==0.6.0; extra == "locked"
@@ -71,6 +72,7 @@ Requires-Dist: langchain-community==0.2.0; extra == "locked"
 Requires-Dist: langchain-core==0.2.0; extra == "locked"
 Requires-Dist: langchain-text-splitters==0.2.0; extra == "locked"
 Requires-Dist: langsmith==0.1.45; extra == "locked"
+Requires-Dist: lingua-language-detector==2.0.2; python_version < "3.13" and extra == "locked"
 Requires-Dist: markdown-it-py==3.0.0; extra == "locked"
 Requires-Dist: marshmallow==3.20.1; extra == "locked"
 Requires-Dist: mdurl==0.1.2; extra == "locked"
@@ -420,9 +422,10 @@ docker run -v <your-config-dir>:/config -p 9527:9527 -e XIAOGPT_HOSTNAME=<your i
 
 注意端口必须映射为与容器内一致，XIAOGPT_HOSTNAME 需要设置为宿主机的 IP 地址，否则小爱无法正常播放语音。
 
-## 推荐的 fork
+## 推荐的类似项目
 
 - [XiaoBot](https://github.com/longbai/xiaobot) -> Go语言版本的Fork, 带支持不同平台的UI
+- [MiGPT](https://github.com/idootop/mi-gpt) -> Node.js 版，支持流式响应和长短期记忆
 
 ## 感谢
 

{xiaogpt-2.82 → xiaogpt-2.84}/README.md

@@ -306,9 +306,10 @@ docker run -v <your-config-dir>:/config -p 9527:9527 -e XIAOGPT_HOSTNAME=<your i
 
 注意端口必须映射为与容器内一致，XIAOGPT_HOSTNAME 需要设置为宿主机的 IP 地址，否则小爱无法正常播放语音。
 
-## 推荐的 fork
+## 推荐的类似项目
 
 - [XiaoBot](https://github.com/longbai/xiaobot) -> Go语言版本的Fork, 带支持不同平台的UI
+- [MiGPT](https://github.com/idootop/mi-gpt) -> Node.js 版，支持流式响应和长短期记忆
 
 ## 感谢
 

{xiaogpt-2.82 → xiaogpt-2.84}/pyproject.toml

@@ -28,9 +28,10 @@ dependencies = [
     "groq>=0.5.0",
     "pyyaml>=6.0.1",
     "langchain-community>=0.0.38",
+    "lingua-language-detector>=2.0.2; python_version < \"3.13\"",
 ]
 dynamic = []
-version = "2.82"
+version = "2.84"
 
 [project.license]
 text = "MIT"
@@ -89,6 +90,7 @@ locked = [
     "langchain-core==0.2.0",
     "langchain-text-splitters==0.2.0",
     "langsmith==0.1.45",
+    "lingua-language-detector==2.0.2 ; python_version < \"3.13\"",
     "markdown-it-py==3.0.0",
     "marshmallow==3.20.1",
     "mdurl==0.1.2",

{xiaogpt-2.82 → xiaogpt-2.84}/xiaogpt/tts/base.py

@@ -118,7 +118,7 @@ class AudioFileTTS(TTS):
                     continue
             logger.debug("Playing URL %s (%s seconds)", url, duration)
             await asyncio.gather(
-                self.mina_service.play_by_url(self.device_id, url),
+                self.mina_service.play_by_url(self.device_id, url, _type=1),
                 self.wait_for_duration(duration),
             )
         await task

{xiaogpt-2.82 → xiaogpt-2.84}/xiaogpt/utils.py

@@ -5,11 +5,14 @@ import os
 import re
 import socket
 from http.cookies import SimpleCookie
-from typing import AsyncIterator
+from typing import TYPE_CHECKING, AsyncIterator
 from urllib.parse import urlparse
 
 from requests.utils import cookiejar_from_dict
 
+if TYPE_CHECKING:
+    from lingua import LanguageDetector
+
 
 ### HELP FUNCTION ###
 def parse_cookie_string(cookie_string):
@@ -69,3 +72,21 @@ def get_hostname() -> str:
     with socket.socket(socket.AF_INET, socket.SOCK_DGRAM) as s:
         s.connect(("8.8.8.8", 80))
         return s.getsockname()[0]
+
+
+def _get_detector() -> LanguageDetector | None:
+    try:
+        from lingua import LanguageDetectorBuilder
+    except ImportError:
+        return None
+    return LanguageDetectorBuilder.from_all_spoken_languages().build()
+
+
+_detector = _get_detector()
+
+
+def detect_language(text: str) -> str:
+    if _detector is None:
+        return "zh"  # default to Chinese if langdetect module is not available
+    lang = _detector.detect_language_of(text)
+    return lang.iso_code_639_1.name.lower() if lang is not None else "zh"

{xiaogpt-2.82 → xiaogpt-2.84}/xiaogpt/xiaogpt.py

@@ -24,9 +24,7 @@ from xiaogpt.config import (
     Config,
 )
 from xiaogpt.tts import TTS, MiTTS, TetosTTS
-from xiaogpt.utils import (
-    parse_cookie_string,
-)
+from xiaogpt.utils import detect_language, parse_cookie_string
 
 EOF = object()
 
@@ -325,7 +323,7 @@ class MiGPT:
         # WTF xiaomi api
         is_playing = (
             json.loads(playing_info.get("data", {}).get("info", "{}")).get("status", -1)
-            >= 1
+            == 1
         )
         return is_playing
 
@@ -333,7 +331,7 @@ class MiGPT:
         is_playing = await self.get_if_xiaoai_is_playing()
         if is_playing:
             # stop it
-            await self.mina_service.player_stop(self.device_id)
+            await self.mina_service.player_pause(self.device_id)
 
     async def wakeup_xiaoai(self):
         return await miio_command(
@@ -390,11 +388,6 @@ class MiGPT:
                 query = f"{query}，{self.config.prompt}"
             # some model can not detect the language code, so we need to add it
 
-            if self.config.tts != "mi":  # mi only say Chinese
-                query += (
-                    "，并用本段话的language code作为开头，用|分隔，如：en-US|你好……"
-                )
-
             if self.config.mute_xiaoai:
                 await self.stop_if_xiaoai_is_playing()
             else:
@@ -420,18 +413,11 @@ class MiGPT:
                 await self.wakeup_xiaoai()
 
     async def speak(self, text_stream: AsyncIterator[str]) -> None:
-        text = await text_stream.__anext__()
-        # See if the first part contains language code(e.g. en-US|Hello world)
-        lang, _, first_chunk = text.rpartition("|")
-        if len(lang) > 7:
-            # It is not a legal language code, discard it
-            lang, first_chunk = "", text
-
-        lang = (
-            matches[0]
-            if (matches := re.findall(r"([a-z]{2}-[A-Z]{2})", lang))
-            else "zh-CN"
-        )
+        first_chunk = await text_stream.__anext__()
+        # Detect the language from the first chunk
+        # Add suffix '-' because tetos expects it to exist when selecting voices
+        # however, the nation code is never used.
+        lang = detect_language(first_chunk) + "-"
 
         async def gen():  # reconstruct the generator
             yield first_chunk